QVAC

QVAC vs llama.rn

QVAC

Open-source ecosystem for local-first peer-to-peer AI on every platform.

llama.rn

React Native bindings for llama.cpp, for iOS and Android.

Key differences

llama.rn and QVAC both run models on phone hardware from JavaScript, and both sit on llama.cpp: llama.rn binds it directly for React Native, and QVAC's Fabric engine is a fork of it. llama.rn covers iOS and Android and works with Expo through expo-build-properties. QVAC covers those two and adds Node.js, Bare, macOS, Windows and Linux.

Task coverage differs. llama.rn provides completion, embeddings, reranking, multimodal input and an experimental neural TTS path through codec.cpp. QVAC adds speech recognition, machine translation through the Bergamot engine from Mozilla, OCR, image generation and RAG, each with models in a registry.

The surfaces differ in width. llama.rn targets React Native and stops there, which keeps it small and close to the engine. The same QVAC code runs on phones, on desktops and in a server process, so a mobile application and its desktop counterpart can share one codebase.

This page compares llama.rn v0.12.9, released 4 August 2026, against QVAC 0.18.2, meaning the SDK together with the Fabric inference engine at v10297.1.1. Every row was checked against the project's own documentation and release notes on 4 September 2026. Both projects move quickly, so check the current release before you make a decision on either one.

Feature matrix

Feature

QVAC

llama.rn

PLATFORMS

macOS

Yes
No

Windows

Yes
No

Linux

Yes
No

Android

Yes
Yes

iOS

Yes
Yes

AI TASKS

Text generation

Yes
Yes

Transcription

Yes

Audio models

Translation

Yes
No

Image generation

Yes
No

OCR

Yes
No

Text-to-speech

Yes

Experimental

RUNTIME SUPPORT

Node.js

Yes
No

Bare

Yes
No

Expo

Yes
Yes

HTTP server

Yes
No

CLI

Yes
No

P2P

Peer discovery

Yes
No

Inference delegation

Yes
No

Encrypted transport

Yes
No

MOBILE SUPPORT

On-device inference

Yes
Yes

LoRA fine-tuning on mobile

Yes
No

Mobile SDK

Yes
Yes

LICENSING

License

Apache 2.0

MIT

Open weights tooling

Yes
Yes

When to choose QVAC

You need speech, OCR, translation or images alongside text.

The same code has to run on desktop or a server.

You want fine-tuning to run on the user's phone.

You would rather call a model registry than manage GGUF files.

When to choose llama.rn

You ship React Native only, and only need an LLM.

You want a thin wrapper you can read end to end.

You want to track llama.cpp releases closely yourself.

You already have model downloading and caching.

Compare

Ready to build with QVAC?

One SDK, every platform, no rent. Grab it and ship your own local-first AI.

npm install @qvac/sdk
Get the SDKRead the docs